tinygrad host for iOS
Run tinygrad code using your iPhone or iPad's GPU
- install and open this app on your iPhone or iPad
- install this tinygrad fork (new_ios branch) on a computer connected to the same wifi network
git clone -b new_ios https://github.com/roryclear/tinygrad.git
cd tinygrad
pip install -e .
- run any tinygrad code
DEV=IOS IP={your iphone/ipad IP address}:6667 PYTHONPATH=. python3.11 examples/gpt2.py --model_size=gpt2Also try YOLOv8 on tinygrad, by caching tinygrad remote batches, models can be easily ran locally.
- Max allowed app RAM < total RAM of iOS device.
- Metal cannot be ran in the background on iOS, the app must be open to run.
| fp32 Matmul size | CoreML time | Tinygrad time | Tinygrad (BEAM=2) time |
|---|---|---|---|
1x1 @ 1x1 |
0.00013 | 5.9e-06 (0.047x) | 5.6e-06 (0.045x) |
1x1 @ 1x4 |
4.9e-05 | 7.5e-06 (0.15x) | 7.3e-06 (0.15x) |
1x1 @ 1x16 |
4.6e-05 | 6.5e-06 (0.14x) | 1.7e-05 (0.38x) |
1x1 @ 1x64 |
4.8e-05 | 6.6e-06 (0.14x) | 8.4e-06 (0.17x) |
1x1 @ 1x256 |
4.4e-05 | 8e-06 (0.18x) | 6.5e-06 (0.15x) |
1x1 @ 1x2048 |
7.2e-05 | 8.9e-06 (0.12x) | 6.3e-06 (0.088x) |
1x4 @ 4x1 |
4.3e-05 | 6e-06 (0.14x) | 6.5e-06 (0.15x) |
1x4 @ 4x4 |
4.5e-05 | 2.2e-05 (0.49x) | 7.5e-06 (0.17x) |
1x4 @ 4x16 |
4.7e-05 | 8.7e-06 (0.19x) | 6.2e-06 (0.13x) |
1x4 @ 4x64 |
4.4e-05 | 6.8e-06 (0.15x) | 3.2e-05 (0.73x) |
1x4 @ 4x256 |
4.2e-05 | 7.5e-06 (0.18x) | 8.6e-06 (0.2x) |
1x4 @ 4x2048 |
4.3e-05 | 8.4e-06 (0.19x) | 7.3e-06 (0.17x) |
1x16 @ 16x1 |
4.3e-05 | 6e-06 (0.14x) | 5.8e-06 (0.13x) |
1x16 @ 16x4 |
4.7e-05 | 2.3e-05 (0.49x) | 6.1e-06 (0.13x) |
1x16 @ 16x16 |
4.7e-05 | 2.1e-05 (0.45x) | 6.6e-06 (0.14x) |
1x16 @ 16x64 |
4.5e-05 | 8.8e-06 (0.2x) | 6e-06 (0.13x) |
1x16 @ 16x256 |
4.5e-05 | 1.2e-05 (0.27x) | 6.5e-06 (0.14x) |
1x16 @ 16x2048 |
6.2e-05 | 1.3e-05 (0.21x) | 1e-05 (0.17x) |
1x64 @ 64x1 |
4.5e-05 | 7.5e-06 (0.17x) | 6e-06 (0.13x) |
1x64 @ 64x4 |
4.5e-05 | 7.5e-06 (0.17x) | 6.1e-06 (0.14x) |
1x64 @ 64x16 |
4.2e-05 | 1.1e-05 (0.25x) | 6.3e-06 (0.15x) |
1x64 @ 64x64 |
4.9e-05 | 1.1e-05 (0.22x) | 7.4e-06 (0.15x) |
1x64 @ 64x256 |
5.9e-05 | 1.7e-05 (0.29x) | 1.7e-05 (0.28x) |
1x64 @ 64x2048 |
6e-05 | 2.6e-05 (0.44x) | 2.5e-05 (0.41x) |
1x256 @ 256x1 |
4.4e-05 | 6.9e-06 (0.16x) | 9.4e-06 (0.21x) |
1x256 @ 256x4 |
4.9e-05 | 9.3e-06 (0.19x) | 8e-06 (0.16x) |
1x256 @ 256x16 |
4.7e-05 | 1.8e-05 (0.39x) | 2.3e-05 (0.5x) |
1x256 @ 256x64 |
5.1e-05 | 2e-05 (0.39x) | 1.9e-05 (0.38x) |
1x256 @ 256x256 |
5.3e-05 | 2.3e-05 (0.44x) | 2.1e-05 (0.4x) |
1x256 @ 256x2048 |
0.00016 | 0.00012 (0.7x) | 8.6e-05 (0.52x) |
1x2048 @ 2048x1 |
4.1e-05 | 1.9e-05 (0.47x) | 9.7e-06 (0.24x) |
1x2048 @ 2048x4 |
5.3e-05 | 2.2e-05 (0.42x) | 1.1e-05 (0.21x) |
1x2048 @ 2048x16 |
5.6e-05 | 0.00012 (2.2x) | 2.8e-05 (0.5x) |
1x2048 @ 2048x64 |
0.00012 | 0.00011 (0.95x) | 4.9e-05 (0.42x) |
1x2048 @ 2048x256 |
0.00016 | 0.00017 (1.1x) | 8e-05 (0.49x) |
1x2048 @ 2048x2048 |
0.0011 | 0.00056 (0.49x) | 0.00055 (0.49x) |
4x1 @ 1x1 |
5.9e-05 | 7.5e-06 (0.13x) | 7.5e-06 (0.13x) |
4x1 @ 1x4 |
4.8e-05 | 7e-06 (0.14x) | 5.4e-06 (0.11x) |
4x1 @ 1x16 |
4.8e-05 | 6.5e-06 (0.14x) | 5.7e-06 (0.12x) |
4x1 @ 1x64 |
4.9e-05 | 5.8e-06 (0.12x) | 5.1e-06 (0.1x) |
4x1 @ 1x256 |
4.7e-05 | 6e-06 (0.13x) | 8e-06 (0.17x) |
4x1 @ 1x2048 |
5.2e-05 | 7.5e-06 (0.15x) | 6.5e-06 (0.13x) |
4x4 @ 4x1 |
4.2e-05 | 6.3e-06 (0.15x) | 6e-06 (0.14x) |
4x4 @ 4x4 |
4.2e-05 | 6.7e-06 (0.16x) | 5.4e-06 (0.13x) |
4x4 @ 4x16 |
3.8e-05 | 8.5e-06 (0.22x) | 6.8e-06 (0.18x) |
4x4 @ 4x64 |
4.6e-05 | 5.5e-06 (0.12x) | 5.8e-06 (0.13x) |
4x4 @ 4x256 |
4.2e-05 | 7.6e-06 (0.18x) | 1.6e-05 (0.38x) |
4x4 @ 4x2048 |
5.5e-05 | 1.3e-05 (0.23x) | 1.7e-05 (0.31x) |
4x16 @ 16x1 |
4.4e-05 | 5.9e-06 (0.13x) | 6.6e-06 (0.15x) |
4x16 @ 16x4 |
3.9e-05 | 6.8e-06 (0.17x) | 1.5e-05 (0.38x) |
4x16 @ 16x16 |
4.3e-05 | 6.9e-06 (0.16x) | 2.2e-05 (0.5x) |
4x16 @ 16x64 |
4.8e-05 | 8.1e-06 (0.17x) | 1.6e-05 (0.34x) |
4x16 @ 16x256 |
4.5e-05 | 1.8e-05 (0.39x) | 7.6e-06 (0.17x) |
4x16 @ 16x2048 |
7e-05 | 2e-05 (0.28x) | 1.2e-05 (0.17x) |
4x64 @ 64x1 |
4.2e-05 | 6.5e-06 (0.15x) | 9.1e-06 (0.22x) |
4x64 @ 64x4 |
4.3e-05 | 7.1e-06 (0.16x) | 8.8e-06 (0.2x) |
4x64 @ 64x16 |
4.3e-05 | 7.5e-06 (0.17x) | 2.1e-05 (0.48x) |
4x64 @ 64x64 |
4.6e-05 | 1.6e-05 (0.34x) | 1.2e-05 (0.26x) |
4x64 @ 64x256 |
4.8e-05 | 2.8e-05 (0.58x) | 1.5e-05 (0.3x) |
4x64 @ 64x2048 |
6.6e-05 | 3.3e-05 (0.5x) | 3.2e-05 (0.48x) |
4x256 @ 256x1 |
4.1e-05 | 8.1e-06 (0.2x) | 6.7e-06 (0.16x) |
4x256 @ 256x4 |
3.9e-05 | 1.1e-05 (0.27x) | 7.7e-06 (0.2x) |
4x256 @ 256x16 |
4.5e-05 | 1.2e-05 (0.27x) | 1.1e-05 (0.24x) |
4x256 @ 256x64 |
4.8e-05 | 3.6e-05 (0.75x) | 1.9e-05 (0.39x) |
4x256 @ 256x256 |
5.3e-05 | 0.00015 (2.8x) | 3.7e-05 (0.7x) |
4x256 @ 256x2048 |
0.00018 | 0.00034 (1.9x) | 0.0001 (0.57x) |
4x2048 @ 2048x1 |
5.2e-05 | 1.8e-05 (0.34x) | 1.1e-05 (0.2x) |
4x2048 @ 2048x4 |
5e-05 | 3.4e-05 (0.68x) | 3.2e-05 (0.64x) |
4x2048 @ 2048x16 |
5.3e-05 | 8.6e-05 (1.6x) | 3.9e-05 (0.74x) |
4x2048 @ 2048x64 |
6.6e-05 | 0.00039 (5.9x) | 6.3e-05 (0.95x) |
4x2048 @ 2048x256 |
0.00017 | 0.0017 (9.7x) | 0.00014 (0.81x) |
4x2048 @ 2048x2048 |
0.0016 | 0.00063 (0.41x) | 0.0015 (0.93x) |
16x1 @ 1x1 |
5.5e-05 | 6.1e-06 (0.11x) | 6.8e-06 (0.12x) |
16x1 @ 1x4 |
4.3e-05 | 9e-06 (0.21x) | 2e-05 (0.47x) |
16x1 @ 1x16 |
4.4e-05 | 6.1e-06 (0.14x) | 5.8e-06 (0.13x) |
16x1 @ 1x64 |
4.7e-05 | 1.5e-05 (0.32x) | 5.9e-06 (0.13x) |
16x1 @ 1x256 |
4.6e-05 | 7e-06 (0.15x) | 5.8e-06 (0.13x) |
16x1 @ 1x2048 |
5.6e-05 | 9.8e-06 (0.17x) | 1.2e-05 (0.21x) |
16x4 @ 4x1 |
4.6e-05 | 1.7e-05 (0.37x) | 1e-05 (0.22x) |
16x4 @ 4x4 |
4.4e-05 | 5.9e-06 (0.13x) | 5.8e-06 (0.13x) |
16x4 @ 4x16 |
4.2e-05 | 1.8e-05 (0.44x) | 9.4e-06 (0.22x) |
16x4 @ 4x64 |
4.8e-05 | 6e-06 (0.13x) | 1e-05 (0.21x) |
16x4 @ 4x256 |
4.6e-05 | 7.7e-06 (0.17x) | 1.5e-05 (0.32x) |
16x4 @ 4x2048 |
6.2e-05 | 1.1e-05 (0.18x) | 2.2e-05 (0.36x) |
16x16 @ 16x1 |
4.2e-05 | 6.6e-06 (0.16x) | 7.2e-06 (0.17x) |
16x16 @ 16x4 |
4.1e-05 | 6.4e-06 (0.16x) | 6.2e-06 (0.15x) |
16x16 @ 16x16 |
4.1e-05 | 8.3e-06 (0.2x) | 6.6e-06 (0.16x) |
16x16 @ 16x64 |
4.1e-05 | 1.6e-05 (0.4x) | 6.8e-06 (0.17x) |
16x16 @ 16x256 |
5.2e-05 | 1.8e-05 (0.35x) | 8.6e-06 (0.17x) |
16x16 @ 16x2048 |
6.5e-05 | 2.4e-05 (0.37x) | 2.5e-05 (0.38x) |
16x64 @ 64x1 |
4.1e-05 | 5.9e-06 (0.14x) | 7.2e-06 (0.18x) |
16x64 @ 64x4 |
4.1e-05 | 1.7e-05 (0.41x) | 1.9e-05 (0.46x) |
16x64 @ 64x16 |
4.5e-05 | 1.9e-05 (0.43x) | 1.1e-05 (0.25x) |
16x64 @ 64x64 |
4.4e-05 | 2.9e-05 (0.67x) | 1.3e-05 (0.29x) |
16x64 @ 64x256 |
5e-05 | 3.8e-05 (0.76x) | 2e-05 (0.4x) |
16x64 @ 64x2048 |
6.1e-05 | 4.2e-05 (0.69x) | 7.7e-05 (1.3x) |
16x256 @ 256x1 |
4.7e-05 | 7.9e-06 (0.17x) | 1.1e-05 (0.23x) |
16x256 @ 256x4 |
4.8e-05 | 1.6e-05 (0.34x) | 2e-05 (0.42x) |
16x256 @ 256x16 |
5e-05 | 3.7e-05 (0.74x) | 1e-05 (0.2x) |
16x256 @ 256x64 |
5.1e-05 | 0.00014 (2.8x) | 1.7e-05 (0.33x) |
16x256 @ 256x256 |
5.3e-05 | 7.3e-05 (1.4x) | 5.4e-05 (1x) |
16x256 @ 256x2048 |
0.00017 | 0.00019 (1.1x) | 0.00023 (1.4x) |
16x2048 @ 2048x1 |
5.8e-05 | 1.9e-05 (0.32x) | 1.6e-05 (0.27x) |
16x2048 @ 2048x4 |
5.5e-05 | 4.7e-05 (0.85x) | 4.1e-05 (0.75x) |
16x2048 @ 2048x16 |
6.1e-05 | 0.00042 (6.9x) | 3e-05 (0.49x) |
16x2048 @ 2048x64 |
6.7e-05 | 0.0017 (25x) | 7.2e-05 (1.1x) |
16x2048 @ 2048x256 |
0.00018 | 0.00048 (2.6x) | 0.00031 (1.7x) |
16x2048 @ 2048x2048 |
0.0015 | 0.0015 (1x) | 0.0013 (0.88x) |
64x1 @ 1x1 |
4.9e-05 | 6.1e-06 (0.12x) | 7.4e-06 (0.15x) |
64x1 @ 1x4 |
4.5e-05 | 1.2e-05 (0.27x) | 5.9e-06 (0.13x) |
64x1 @ 1x16 |
4.4e-05 | 6.8e-06 (0.16x) | 5.8e-06 (0.13x) |
64x1 @ 1x64 |
4.9e-05 | 7.5e-06 (0.15x) | 8.6e-06 (0.18x) |
64x1 @ 1x256 |
4.5e-05 | 9.3e-06 (0.21x) | 8.6e-06 (0.19x) |
64x1 @ 1x2048 |
0.00011 | 3.2e-05 (0.28x) | 1.7e-05 (0.15x) |
64x4 @ 4x1 |
4.8e-05 | 7e-06 (0.14x) | 7.6e-06 (0.16x) |
64x4 @ 4x4 |
4.5e-05 | 6.5e-06 (0.15x) | 1.9e-05 (0.43x) |
64x4 @ 4x16 |
4.9e-05 | 1.6e-05 (0.33x) | 5.9e-06 (0.12x) |
64x4 @ 4x64 |
4.5e-05 | 9.1e-06 (0.2x) | 7.8e-06 (0.17x) |
64x4 @ 4x256 |
5.2e-05 | 1.1e-05 (0.21x) | 2.1e-05 (0.41x) |
64x4 @ 4x2048 |
8.4e-05 | 3.6e-05 (0.42x) | 2.4e-05 (0.29x) |
64x16 @ 16x1 |
4.1e-05 | 1.9e-05 (0.47x) | 6.7e-06 (0.16x) |
64x16 @ 16x4 |
4e-05 | 8.8e-06 (0.22x) | 7.1e-06 (0.18x) |
64x16 @ 16x16 |
4.1e-05 | 1.6e-05 (0.4x) | 1.6e-05 (0.39x) |
64x16 @ 16x64 |
4.2e-05 | 2.1e-05 (0.49x) | 1e-05 (0.25x) |
64x16 @ 16x256 |
4.8e-05 | 3e-05 (0.64x) | 1.8e-05 (0.38x) |
64x16 @ 16x2048 |
0.0001 | 6.1e-05 (0.6x) | 6.2e-05 (0.61x) |
64x64 @ 64x1 |
4.8e-05 | 8.9e-06 (0.18x) | 7.7e-06 (0.16x) |
64x64 @ 64x4 |
4.3e-05 | 1.1e-05 (0.25x) | 1.9e-05 (0.44x) |
64x64 @ 64x16 |
4.2e-05 | 3.3e-05 (0.78x) | 2.1e-05 (0.5x) |
64x64 @ 64x64 |
4.7e-05 | 2.6e-05 (0.55x) | 2e-05 (0.42x) |
64x64 @ 64x256 |
5.1e-05 | 4.6e-05 (0.9x) | 4.9e-05 (0.97x) |
64x64 @ 64x2048 |
7.9e-05 | 0.00014 (1.8x) | 0.0003 (3.7x) |
64x256 @ 256x1 |
4.6e-05 | 1.5e-05 (0.33x) | 1.1e-05 (0.23x) |
64x256 @ 256x4 |
5.2e-05 | 4e-05 (0.76x) | 1.2e-05 (0.22x) |
64x256 @ 256x16 |
5e-05 | 0.00017 (3.3x) | 1.8e-05 (0.36x) |
64x256 @ 256x64 |
5.2e-05 | 7.5e-05 (1.4x) | 3.5e-05 (0.68x) |
64x256 @ 256x256 |
5.4e-05 | 8.8e-05 (1.6x) | 0.0001 (1.9x) |
64x256 @ 256x2048 |
0.00018 | 0.00049 (2.8x) | 0.00056 (3.1x) |
64x2048 @ 2048x1 |
7.5e-05 | 3.5e-05 (0.47x) | 2.7e-05 (0.36x) |
64x2048 @ 2048x4 |
8e-05 | 0.00015 (1.9x) | 4.3e-05 (0.54x) |
64x2048 @ 2048x16 |
8.8e-05 | 0.0017 (19x) | 7.6e-05 (0.87x) |
64x2048 @ 2048x64 |
8.5e-05 | 0.0006 (7.1x) | 0.00027 (3.1x) |
64x2048 @ 2048x256 |
0.00023 | 0.0007 (3x) | 0.00077 (3.3x) |
64x2048 @ 2048x2048 |
0.0027 | 0.0039 (1.4x) | 0.0041 (1.5x) |
256x1 @ 1x1 |
4.9e-05 | 7.5e-06 (0.15x) | 7.9e-06 (0.16x) |
256x1 @ 1x4 |
4.3e-05 | 7e-06 (0.16x) | 8.5e-06 (0.2x) |
256x1 @ 1x16 |
4.4e-05 | 7.8e-06 (0.18x) | 6.7e-06 (0.15x) |
256x1 @ 1x64 |
4.6e-05 | 9.7e-06 (0.21x) | 7e-06 (0.15x) |
256x1 @ 1x256 |
6.2e-05 | 1.9e-05 (0.31x) | 1.4e-05 (0.22x) |
256x1 @ 1x2048 |
0.00015 | 0.00011 (0.73x) | 7.9e-05 (0.53x) |
256x4 @ 4x1 |
4e-05 | 7.9e-06 (0.2x) | 7.2e-06 (0.18x) |
256x4 @ 4x4 |
4.2e-05 | 7.5e-06 (0.18x) | 7e-06 (0.17x) |
256x4 @ 4x16 |
4.8e-05 | 1e-05 (0.21x) | 6.4e-06 (0.13x) |
256x4 @ 4x64 |
4.7e-05 | 1.2e-05 (0.25x) | 1.1e-05 (0.23x) |
256x4 @ 4x256 |
5.8e-05 | 2.1e-05 (0.35x) | 2.2e-05 (0.38x) |
256x4 @ 4x2048 |
0.00017 | 0.00012 (0.71x) | 0.00012 (0.69x) |
256x16 @ 16x1 |
4.8e-05 | 8.5e-06 (0.18x) | 7.5e-06 (0.16x) |
256x16 @ 16x4 |
4.7e-05 | 2.7e-05 (0.58x) | 1.1e-05 (0.23x) |
256x16 @ 16x16 |
5.1e-05 | 1.6e-05 (0.31x) | 8.4e-06 (0.16x) |
256x16 @ 16x64 |
5.3e-05 | 1.7e-05 (0.32x) | 2.3e-05 (0.44x) |
256x16 @ 16x256 |
6.7e-05 | 3.7e-05 (0.55x) | 4.1e-05 (0.61x) |
256x16 @ 16x2048 |
0.00024 | 0.0002 (0.84x) | 0.00038 (1.6x) |
256x64 @ 64x1 |
4.5e-05 | 9.3e-06 (0.21x) | 9e-06 (0.2x) |
256x64 @ 64x4 |
4.6e-05 | 2.8e-05 (0.6x) | 2.2e-05 (0.48x) |
256x64 @ 64x16 |
5.4e-05 | 3.5e-05 (0.65x) | 1.7e-05 (0.32x) |
256x64 @ 64x64 |
4.9e-05 | 3.9e-05 (0.8x) | 5.3e-05 (1.1x) |
256x64 @ 64x256 |
6e-05 | 7.5e-05 (1.3x) | 0.00013 (2.2x) |
256x64 @ 64x2048 |
0.00016 | 0.00051 (3.1x) | 0.00053 (3.2x) |
256x256 @ 256x1 |
6.4e-05 | 2.1e-05 (0.33x) | 1.7e-05 (0.27x) |
256x256 @ 256x4 |
6.1e-05 | 8.1e-05 (1.3x) | 5.5e-05 (0.9x) |
256x256 @ 256x16 |
7.1e-05 | 9.6e-05 (1.4x) | 5.6e-05 (0.78x) |
256x256 @ 256x64 |
5.9e-05 | 8.9e-05 (1.5x) | 0.00016 (2.7x) |
256x256 @ 256x256 |
7e-05 | 0.00026 (3.7x) | 0.00028 (4x) |
256x256 @ 256x2048 |
0.00035 | 0.0016 (4.5x) | 0.0029 (8.3x) |
256x2048 @ 2048x1 |
0.00025 | 0.0001 (0.4x) | 8.4e-05 (0.33x) |
256x2048 @ 2048x4 |
0.00025 | 0.00057 (2.3x) | 0.00026 (1.1x) |
256x2048 @ 2048x16 |
0.00028 | 0.00072 (2.6x) | 0.00052 (1.8x) |
256x2048 @ 2048x64 |
0.00028 | 0.00061 (2.1x) | 0.00073 (2.6x) |
256x2048 @ 2048x256 |
0.0011 | 0.0021 (1.9x) | 0.0029 (2.7x) |
256x2048 @ 2048x2048 |
0.0057 | 0.0069 (1.2x) | 0.011 (1.9x) |
2048x1 @ 1x1 |
5e-05 | 8e-06 (0.16x) | 9.2e-06 (0.18x) |
2048x1 @ 1x4 |
5.4e-05 | 8.3e-06 (0.15x) | 6.6e-06 (0.12x) |
2048x1 @ 1x16 |
5.3e-05 | 1.3e-05 (0.25x) | 7.9e-06 (0.15x) |
2048x1 @ 1x64 |
7.3e-05 | 3.1e-05 (0.43x) | 2.6e-05 (0.35x) |
2048x1 @ 1x256 |
0.00016 | 0.00011 (0.7x) | 7.3e-05 (0.47x) |
2048x1 @ 1x2048 |
0.0024 | 0.00092 (0.38x) | 0.0007 (0.29x) |
2048x4 @ 4x1 |
5.2e-05 | 8.8e-06 (0.17x) | 8.1e-06 (0.16x) |
2048x4 @ 4x4 |
5.3e-05 | 2.6e-05 (0.48x) | 9.2e-06 (0.17x) |
2048x4 @ 4x16 |
5.4e-05 | 1.5e-05 (0.28x) | 1.2e-05 (0.21x) |
2048x4 @ 4x64 |
8e-05 | 3.6e-05 (0.45x) | 3.1e-05 (0.39x) |
2048x4 @ 4x256 |
0.00017 | 0.00012 (0.74x) | 7.9e-05 (0.48x) |
2048x4 @ 4x2048 |
0.0024 | 0.001 (0.41x) | 0.00079 (0.33x) |
2048x16 @ 16x1 |
5.6e-05 | 2.7e-05 (0.48x) | 1.2e-05 (0.21x) |
2048x16 @ 16x4 |
5.9e-05 | 2.6e-05 (0.44x) | 1.8e-05 (0.3x) |
2048x16 @ 16x16 |
6.4e-05 | 3.4e-05 (0.53x) | 1.9e-05 (0.3x) |
2048x16 @ 16x64 |
0.0001 | 6e-05 (0.6x) | 0.00011 (1.1x) |
2048x16 @ 16x256 |
0.00023 | 0.0002 (0.87x) | 0.00035 (1.5x) |
2048x16 @ 16x2048 |
0.0028 | 0.0015 (0.53x) | 0.001 (0.38x) |
2048x64 @ 64x1 |
8e-05 | 3e-05 (0.37x) | 2.5e-05 (0.31x) |
2048x64 @ 64x4 |
8.1e-05 | 5e-05 (0.62x) | 6.5e-05 (0.8x) |
2048x64 @ 64x16 |
9.9e-05 | 9.9e-05 (1x) | 0.00013 (1.3x) |
2048x64 @ 64x64 |
7.6e-05 | 0.00014 (1.8x) | 0.00035 (4.6x) |
2048x64 @ 64x256 |
0.00015 | 0.0005 (3.3x) | 0.00087 (5.7x) |
2048x64 @ 64x2048 |
0.0024 | 0.0039 (1.7x) | 0.0045 (1.9x) |
2048x256 @ 256x1 |
0.00026 | 0.00012 (0.44x) | 8e-05 (0.3x) |
2048x256 @ 256x4 |
0.00025 | 0.00013 (0.53x) | 0.00015 (0.61x) |
2048x256 @ 256x16 |
0.00029 | 0.00038 (1.3x) | 0.00024 (0.81x) |
2048x256 @ 256x64 |
0.00023 | 0.00049 (2.2x) | 0.00075 (3.3x) |
2048x256 @ 256x256 |
0.00042 | 0.0019 (4.5x) | 0.0029 (6.9x) |
2048x256 @ 256x2048 |
0.0089 | 0.012 (1.3x) | 0.0096 (1.1x) |
2048x2048 @ 2048x1 |
0.0017 | 0.0011 (0.66x) | 0.00056 (0.33x) |
2048x2048 @ 2048x4 |
0.0018 | 0.00095 (0.54x) | 0.00097 (0.55x) |
2048x2048 @ 2048x16 |
0.0019 | 0.003 (1.6x) | 0.0026 (1.4x) |
2048x2048 @ 2048x64 |
0.0035 | 0.0039 (1.1x) | 0.0034 (0.99x) |
2048x2048 @ 2048x256 |
0.012 | 0.008 (0.67x) | 0.02 (1.7x) |
2048x2048 @ 2048x2048 |
0.021 | 0.036 (1.7x) | 0.034 (1.6x) |